The specification of which terms to linearise is messy

[tommbendall]

Currently the terms to be linearised in an equation are specified through a dictionary, pairing term labels and fields. e.g.

    def __init__(self, state, family, degree, Omega=None, sponge=None,
                 extra_terms=None,
                 terms_to_linearise={'u': [time_derivative],
                                     'rho': [time_derivative, transport],
                                     'theta': [time_derivative, transport]},
                 u_transport_option="vector_invariant_form",
                 diffusion_options=None,
                 no_normal_flow_bc_ids=None,
                 active_tracers=None):

Would it be better to replace terms_to_linearise with a term_filter that gets used by a label_map? How do we easily make this configurable?

[tommbendall]

This is handled much more neatly following #317